最有看点的互联网金融门户

最有看点的互联网金融门户
国际资讯

P2P贷款质量分析:拖欠行为与违约率无必然联系

在p2p借贷中,投资者可以利用多种征信数据对其投资战略进行评估。征信机构收集的行为数据种类繁多,比如历史拖欠行为。所谓“拖欠行为”是指未能按时履行最低还款义务的行为。何时向征信机构上报拖欠记录由各个贷款机构自己的内部协议决定。一些贷款机构仅在借款人发生至少两次拖欠行为时才会向征信机构上报一次拖欠记录,其他的贷款机构则是发现一次上报一次。因此,不能将所有拖欠行为同等对待。有些拖欠行为仅是拖沓支付行为,而非借款人资金困难。Prosper公司的数据总结了过去七年中全部拖欠账户的历史拖欠变量。

要了解这一变量的可预测性,我们可以先来看一下2012年Prosper公司发放的贷款情况。我们按贷款信用等级表示下列分布情况,其中一些贷款有拖欠行为,另外一些无拖欠行为。正如所预期的那样,信用风险等级越高,曾经发生拖欠贷款行为的比例也就越高。2012年,共有30%的借款人发生至少一次历史拖欠行为。从该术语的广义角度考虑,这一结果并不出人意料。况且七年时间很长,不排除在此期间借款人经济状况发生巨大变化的可能。

Prosper_Delq_0 第1张

接下来的问题是,有拖欠行为相对无拖欠行为是否改变了总量。下面的柱形图显示了这两组之间在违约率(鉴于2012年的贷款尚未全部到期,我们所收集的是至少发生过一次逾期支付的违约行为)上的细微差别。

Prosper_Delq_1 第2张

按信用等级分布的拖欠记录也可以看到同样的细微差别模型,尽管我们看到,风险等级越高,两组间的差异越不明显,甚至出现拖欠行为减少的可能。

Prosper-Delq-2 第3张

到目前为止,我们将每一次拖欠记录集中在一起来看这些账户。许多借款人都可能有一到两次拖欠记录,如果不只是少数几次拖欠记录,该变量才可能用来预测未来的违约行为。

USE-USE-USE 第4张

此处,我们使用征信机构局提供的拖欠记录次数来表示2012年Prosper公司借款人的违约率。该柱形图最右侧的柱形条是具有20次以上(含20次)拖欠记录的所有账户,其违约率普遍相似,而且随着拖欠次数的增加,违约情况并未恶化。确切地说,有三次拖欠记录的借款人违约率最高。这一数字极其随意,不大可能代表未来的某种可重复发生模型。从这个角度看,信用分析有时可能就不能简单地用数理分析。详细审查战略中的所有信息,从而确保所见到的模型是真实模型,并且在未来具有可靠性,这些才是重点。至于历史拖欠行为,并不能预测一般类型的违约行为,其中的影响因素有很多:比如拖欠行为的一般特性,进行评估的七年期限,以及进行信用筛查的贷款平台各异。

Investors in online lending have a wealth of credit bureau data to assess when developing an investment strategy.  Credit bureaus track many types of behavior, including delinquencies.  A delinquency is defined as a failure to pay the minimum payment on an obligation by the time it was due.  Credit issuers have their own internal protocols that determine when delinquency data is transmitted to the credit bureaus.  Some issuers may only report a delinquency if the borrower has missed at least two payments, others report after one missed payment.  Because of this, all delinquencies are not the same, and some could be simply associated with sloppy payment behavior, as opposed to financial distress.  Prosper’s data includes a summary historical delinquency variable that contains the total number of delinquent accounts over the last seven years.

To understand if this variable is predictive, we will take a look at Prosper loans made in 2012.  Below, we show the distribution by credit grade of loans with and without a delinquency.  As would be expected, the higher risk credit grades include a higher proportion of loans with previous delinquencies.  In total, about 30% of borrowers from 2012 have at least one historical delinquency.  This is not a surprise given the term’s broad definition.  Also, seven years is a long period of time, one in which a borrower’s financial situation could have drastically changed.

Prosper_Delq_0 第1张

The next question is whether having a delinquency vs. not makes a difference in aggregate.  The below chart shows that there is a small difference in default rate (defined as at least one late payment, which is what we track given that 2012 loans still have not fully matured) between the two groups.

Prosper_Delq_1 第2张

The same pattern exists by credit grade, although we see that it becomes less significant and even potentially reverses in the higher-risk grades.

Prosper-Delq-2 第3张

So far, we’ve been looking at accounts with every delinquency lumped together.  It is possible that many borrowers have one or two delinquencies, and the variable only becomes predictive of future default if there are more than just a few occurrences.

USE-USE-USE 第4张

Here we show the default rate for Prosper 2012 borrowers based on how many delinquencies are on the credit bureau.  The far right bar in the chart includes all those accounts with 20 or more delinquencies.  The default rate is similar across the board, and does not get worse as the number of delinquencies increases.  In fact, borrowers with three delinquencies have one of the highest default rates.  This number is quite arbitrary and unlikely to be representative of a pattern that would repeat in the future.  This kind of insight is where credit analysis goes from being a science to an art.  It is important to scrutinize all inputs to a strategy to ensure the pattern seen is actually a pattern, and makes sense to rely on in the future.  In the case of historical delinquencies, it does not appear to be predictive of default in general.  This is likely because of the general nature of delinquencies, the length of the seven-year time period being assessed, and the initial credit screen performed by the origination platform itself.


用微信扫描可以分享至好友和朋友圈

扫描二维码或搜索微信号“iweiyangx”
关注未央网官方微信公众号,获取互联网金融领域前沿资讯。

发表评论

发表评论

您的评论提交后会进行审核,审核通过的留言会展示在下方留言区域,请耐心等待。

评论

您的个人信息不会被公开,请放心填写! 标记为的是必填项

取消

京东征战P2P 成BATJ中首个明确布局网贷的巨头

三水 | 网贷之家 11-17

赴美IPO热潮下,理性观察上市系平台数据表现

网贷之家 11-17

P2P引爆下一场次贷危机?美联储报告粉碎P2P三大神话

董云峰 | 新金融琅琊... 11-16

未央今日播报:深圳发布P2P防电信诈骗指引 美国要求ICO代币必须登记为证券

未央研究 11-13

P2P老司机经验分享会

七点半理财 | 网贷之家 11-13

版权所有 © 清华大学五道口金融学院互联网金融实验室 | 京ICP备17044750号-1